Non-stationary Models of Software Testing Strategies with Probabilistic Parameters for Fault Detection [Article Withdrawn]
Abstract
Problem definition: The modern software development standards demand to specify the schedule and resources for successful project implementation. An important aspect is providing the predefined quality level of the software. Modeling of software testing helps to plan the resources and final quality at early stages of the project execution. The available software testing models do not take into account the probabilistic nature of error detection and resolution. Usually models are based on a binary classification of software modules which are divided into three classes: potentially faulty, definitely faulty and fault-free ones. Purpose: The existing software testing models should be improved taking into account the probabilistic nature of error detection and resolution. Results: Three dynamic models of processes (strategies) are developed for software testing, using error detection probability for each software module. For each strategy, modified labeled graphs are built, along with differential equation systems and their numerical solutions. This helped to compute probabilistic characteristics of the test processes and states: probability states, distribution functions for fault detection and elimination, mathematical expectations of random variables, amount of the detected or fixed errors. The strategies have been compared by their quality indexes. The proposed models allow us to use the reliability estimates for each individual module. This improves the accuracy of software test process modeling and helps to take into account the viability (power) of the tests. Using these models, we can search for ways to improve software reliability by generating tests which detect errors with the highest probability.